BOLTED OR WELDED ASSEMBLY SIMULATIONS

OVERCOMING SERIOUS OBSTACLES TO ADVANCED FEA VALUE

presented at NAFEMS 2003 World Congress, Orlando, Florida
reprinted 2015 by the author &
former owner of Integra Engineering, Inc.

Keith J. Orgeron, P.E.

now of Orgeron Engineering, PLLC
Spring, TX 77389 731-397-8625
keith.orgeron@OEngineer.com

Abstract
Given that finite element analysis (FEA) value
is relatively proportional to the complexity of
the valid solution obtained, then multi-body
assembly simulations represent great value.
Such models (Figure 1) inherently involve
component contact interaction and typically
other nonlinear geometric or material and
dynamic event parameters.

software, provides the final

establishing the net FEA value.

basis

for

The 7-Step FEA process model is introduced

as a system for performance and validation.
These structured decision trees help to organize
and thus simplify the complex FEA efforts at
hand.

The advanced FEA technology required for

assembly simulations is not only complex but
also remains difficult to wield cost-effectively
toward establishing validity and accuracy.
There are many and serious obstacles in
achieving valid, converged solutions. Often,
several models processed multiple times are
required. It is no wonder that experience,
foresight and adequate resources can be critical
to successfully delivering advanced FEA value.
Most nonlinear solutions require achieving
convergence for the simulations event duration
of interest. Unfortunately, there are literally
hundreds of ways to inadvertently cause
spurious assembly model behavior that can
grind the solution to a halt with excessive
numerical iterations.
This paper attempts to outline the most common
and offensive spurious behaviors defeating
solution convergence. By way of example it
provides general guidance and details several
novel techniques to help the analyst overcome
these serious obstacles.
The case is made that validation of the FEA
work, as opposed to the verification of the FEA

Figure 1 Pressure-test of ball valve assembly,

Mises stresses, 50x

INTRODUCTION
Most manufactured products are assemblies,
subassemblies or parts that makeup an assembly.
Assemblies inherently require joining methods
such as bolting, welding or use of other fasteners
or bonds. Bolted products such as the ball valve
assembly represented by the half-model below
(Figure 2) are probably the most common. (Note
that the threaded holes are all blind.)

graphics to truly increase our efficiency and

potential for accurate simulations.
So, the question remains, Why do so few
engineers/analysts
perform
multi-body
simulations of even the most common
mechanical events, such as bolt pre-loading for
instance? The advanced FEA technology
required for assembly simulations is not only
complex but remains difficult to wield such that
validity and accuracy can be cost-effectively
established to yield valuable results
Hopefully, the system and techniques conveyed
in this paper will help analysts to overcome some
of these serious obstacles to advanced FEA
value.

Figure 2 Valve body & bolt stresses at limit

load, 30x
Industrial
and
commercial
equipment
manufacturers comprise a large or the primary
source of funding for engineering analysis
activities.
Most industrial and commercial
products are candidates for advanced FEA
simulations of at least mechanical or thermomechanical physical events in one or more of the
following areas:

The model in Figure 3 below successfully

addressed the critical bolting preload issues that
prevail in many assembly simulations. This
validation model incorporates several of the
techniques elaborated within this paper to costeffectively refine key parameters prior to
processing the larger model.

Prototype virtual testing, design liability

containment, failure analysis, sensitivity
studies, design optimization, model rating and
sizing, cost reduction, component interactions,
code compliance, preload adjustment,
survival conditions and installation, bolt or
weld residual effects.
FEA, as more than a 50 year old technology, still
appears to favor and focus upon the traditional
method of simulating a single part, regardless of
the inherent and sometimes huge idealization
errors. This is where advanced, nonlinear FEA
excels, in achieving a much better representation
of the physical event idealized, through more
elaborate and often coupled, motion, material,
loading and constraint set representations.
Some of the technology developed 30 years ago
is today not in common use, such as nonlinear
material and geometric solution capabilities.
Then, besides computer hardware improvements,
the last decade has seen the addition of FEA
surface contact capabilities, CAD interfaces,
auto-meshing, fast solvers and high-powered

Figure 3 Eighth-model of bolt-preloaded valve,

150x
This paper seeks to identify the key obstacles in
achieving FEA value for assembly simulations,
focusing on the issues of convergence of a
nonlinear time-history direct integration solution
and validation of the work.
Modeling and validation techniques will be
presented that have been used successfully in
achieving convergence and confirming FEA
value, while cutting net analyst and processing
time expenditures.
The scope of this paper is limited to focusing
upon the modeling, processing, evaluation and

presentation of nonlinear dynamic or pseudostatic time-history simulations employing direct

implicit time integration.
Unless noted,
continuum element use is considered.

features are some of the most critical, the newest

and the least used.

FEA VALUE =
COMPLEXITY + ACCURACY
FEA value is often perceived as the timely
generation of complex analysis results and
conclusions. The presented results are also often
believed to be valid and accurate enough,
but these definitions are highly subjective.
The very nature of undertaking a complex
analysis implies the need for validation in order
to assure both accuracy and relevance of the
models results to the idealizations sought for
satisfying the project objectives. Simply stated,
more complex models have more opportunities
for error.
Yet, in spite of the pitfalls for these errors,
advanced FEA offers more exact, sometimes far
more exact representations of the idealized
problem at hand.
Novice analysts often have initial difficulties
achieving converged solutions, and would be
more challenged to validate their many
decisions. However, these same novices may
also lack the engineering experience to justify
the many simplifying assumptions inherent in the
simpler linear static or linear dynamic analysis.
Though verifying linear loading and constraint
sets is a more common engineering task, the
question of accuracy must address the missing
content of nonlinear behavior. Today, this issue
can best be addressed by comparison with
nonlinear FEA or by physical testing. Figure 4
above illustrates a very nonlinear behavior of the
center bolt as it stretches.

ADVANCED NONLINEAR FEA FOR

ASSEMBLIES
Assembly simulation virtually requires multibody nonlinear contact and interaction for the
most accurate behavior results. Component
contact surfaces provide part or all of the loading
and boundary conditions required to stabilize
most or all components of these FEA models.
Many of the various classes of contact, such as
finite sliding and friction, can only be analyzed
using nonlinear FEA capabilities. These model

Figure 4 Highly nonlinear behavior of center

bolt
Nonlinear geometric solutions are also required
for assembly simulations when they involve
large displacements, rotations or strains. Figure
5 illustrates the valve bonnets nonlinear reaction
behavior as bolt elongation and plasticity cause
load redistribution, clearly requiring a nonlinear
geometric solution. Figure 6 presents some of
the details of the geometry included in the bolted
assembly model, partially detailing the blind
holes.
Assembly simulations involving impact or shock
loading require dynamic solutions coupled with
other nonlinear capabilities, otherwise known as
nonlinear dynamic solutions.

A time history direct integration FEA processor

can readily solve nonlinear dynamic or pseudostatic solutions of nearly any duration.
Larger companies with long histories of FEA use
rely upon their model validation database for
cost-effective support of their analyst decisions.

IDENTIFYING CONVERGENCE &

VALIDATION OBSTACLES
In general, regardless of the number of modeled
components or the complexity of the many
nonlinear features employed for a simulation, the
following philosophy can help guide both the
identification and resolution of convergence
difficulties:
Given that only smaller portions of valid,
nonlinear FEA simulations actually involve
significant nonlinear behavior If each
increment or phase of the event simulated
can be made to behave as linear as the
models idealization, there should be no
convergence difficulties related to spurious
processor calculations, such as surface
contact chatter, yielding or resonant
vibrations
of
non-critical
model
components.

Figure 5 Valve bonnets nonlinear contact

loading behavior (Note: the ordinate label
should read Surface Reaction Force, lbs)

Essentially, all convergence challenges

should be concentrated on the truly critical
model components and their features
having the nonlinear behavior originally
idealized. Phenomena such as surface
impact, plasticity, large displacements or
motion of critical model components will
justifiably demand the most processing
time.

Bolt Preloading & Residual Effects

The primary reason for including a preload of
any kind is to achieve the residual effects that
comprise deflections or distortions, internal
reaction loads and stresses.
Achieving accurate representation of these
residual effects is or should be one of the main
focuses of any assembly simulation model, as in
the case of threaded fasteners.
One study concluded that initial bolt stress
remains a key parameter in the joint
performance [1]
Figure 6 Center-cut valve section Mises
stresses, 30x

Threaded fasteners of any kind represent a

significant modeling challenge, even with the
simpler headed bolt and nut.
Also common but much more difficult to model
are the threaded connection designs for
engagement of male fasteners with blind holes.
This special case is almost nonexistent in FEA
literature.
Figure 7 at right depicts the cast body of the
featured ball valve test assembly model detailed
earlier (Figures 1, 2 & 6). The foreground image
illustrates the preload condition and the
background image, the subsequent pressure
loaded condition.
Both images have their
displacements scaled ~300x for clarity.
Note the extruding and flaring effects achieved at
the visible rims of all threaded holes, the closure
plate seating surface stresses during preload, the
concavity of the preloaded barrel and the
convexity of the pressure-loaded barrel.
This is the basic behavior of such a component.
It is also relatively intuitive, except for the
significance of the hole geometry distortions.
The behavior of these model features may also
significantly affect clearances, seating surfaces
and stress risers.
The inclusion of thread axial and radial stiffness
is critical to achieving such results.
Yet,
modeling of the 3D helical thread or even
individual concentric rings of threads is too
laborious to include in such simulations.
Other more obvious modeling difficulties include
the many options for execution of and validation
of fastener preloading. Bolt shortening by
volumetric or thermal shrinkage is a common
approach.
Part-to-part contact and load transfer without
physically bonding the fastener to its mating
threaded part is necessary to preclude erroneous
tensile and shear effects, especially for blind
threaded connections.
Surface contact of bolt head or nut with their
mating surfaces must not be too soft as to
inadvertently impart a controlling spring into the
connections stiffness system.
Frictional surface contact is intuitively required
for such assemblies, anticipating the friction to
stabilize the sliding tendencies.
But, such
contact features generates additional processing
and collectively more convergence error.

Figure 7 Cast valve body under preload

(foreground)
and
subsequent
pressure
(background), 300x

Earth Entry Events

An Earth entry event such as the onset of
gravity loading is one of the most common
loading obstacles encountered for the more
flexible and larger scale models, especially when
simulating dynamic behavior.
It is often taken for granted in nature that gravity
is constantly acting upon the mass of all objects
whose behavior comprise residual deflections
and loading. Our musculoskeletal system is
good example of this.
Idealization of the physical event must account
for this transient loading as if the model was just
entering the Earths gravitational field.
The problems are that gravity is not
automatically preloaded into the model, and that
there is a need to have sustained gravitational
acceleration prior to initiating other events in the
time history.
Unless many small load steps are used, the
chance of exciting one or more modes of the
structures natural vibrations is quite high.
These vibrations alone can destroy a solutions
convergence, drastically cutting the time step
while attempting to capture each oscillation.

What a shame, when typically this data is of no

interest!

calculations are usually implemented using a

lagrangian or penalty stiffness approach.

For smaller scale models gravity is not

significant relevant to the other load intensities
and is often excluded.

The penalty stiffness approach is considered as

the easier and more stable approach to use,
though the lagrangian methods can be more
accurate. However, the analysts procedure
dominates the error.

Other Body Accelerations

Centrifugal acceleration imparted to a static part,
initial rotation of a dynamic part, or other initial
acceleration required for simulating the event is
often a similar FEA obstacle as well.

Multi-body Contact & Interaction

Establishing multi-body contact between
component parts of an assembly simulation
model is probably the most significant obstacle
of any nonlinear FEA issue.
Surface contact permits the sensitive simulation
of gear dynamics, as shown in Figure 8. Profile
errors effects are evident with acceleration plots.

Algors nonlinear FEA suite provides multi-body

surface contact capability that is implemented
using the penalty stiffness approach.

Surface Contact Chattering

Surface contact chattering is often a spurious
model behavior requiring possibly millions of
contact algorithm calculations yielding results of
no interest to the analyst. This chatter is form of
vibration that can be caused by a lack of surface
smoothness, due to the mesh topology, and/or by
excessive surface contact stiffness.
It can also be caused by inadequate restraint
during contact and allowing excitation of the
free-body or other modes of vibration. This
chattering is can be developed for finite sliding
with friction or for impact events of multi-body
models (either intentioned or unintentional
impact) during initial contact engagement.

Weld Shrinkage & Residual Effects

The simulation of welded assemblies is usually
focused upon the residual effects of the welding
process heat transfer, weld freezing and the
interactions of the welded parts.

Figure 8 Evidence of profile error dynamics

and load sharing for a rack-pinion gear set.
Multi-body contact cannot be established without
assembly component reaction loads and
deflections.
These residual component
displacements cannot usually be imparted as an
input.
Typically, the nonlinear processor attempts to
solve for these residual loads and deflections on
the very first time step of the simulation. Thus,
time step size can be a significant parameter.
The most accurate and flexible nonlinear contact
method to employ is surface contact. Contact

Originally, all parent metal components begin as

cool, loose and virtually unsupported parts. The
weld deposition of molten metal to these parts
injects a thermal shock load to the system and
begins the transient thermo-mechanical event of
joining them together as one.
Assuming that the heat transfer shock loading is
not too extreme to achieve solution convergence,
then the coupled stiffness solution must be able
to carefully support but not over-constrain the
loose parent material as the freezing welds
stiffen and provide their own relative constraints.
Developing physically valid material properties
and event parameters is no small task for this
multi-physics simulation. Typically, there exists
gaps between some parts having no weld metal,
which may require surface contact modeling.

The above issues collectively form a set of major

obstacles to performing welded assembly
simulations. Issues related to appropriate weld
geometry, mesh densities, mesh gradients and
three-dimensional, moving heat load application
to simulate the welding torch further complicates
this advanced FEA work. (See Figure 9.)

establish the computational

validity and accuracy.

model

2.) Project data input, loading and

constraint set are evaluated against the
event idealization and project objectives
to establish validity, accuracy and
relevance of the results.

Contact Friction & Stick-Slip

Use of the nonlinear contact friction capability
can be a very accurate modeling decision, but
under certain conditions it can also be wrought
with untimely and large solution errors causing
havoc with convergence.
The stick-slip friction phenomenon is often
perceived as one bodys acceleration during
slipping over another one, just after a period of
sticking. This is a rather global view of the
event. When, especially for the more flexible
bodies, accelerations and decelerations occur at a
node level.
Imagine each body divided into many parts,
possibly as fine as at each node. For each node,
normal reactions and frictional resistance to
relative motion are calculated.
Obviously,
adjacent nodes are coupled to each other by the
stiffness of their common finite element.
The resulting behavior can be as odd as that of a
caterpillar crawling along the ground, with each
set of legs taking limited small strides in
sequence or in tandem. The author believes that
this is in fact representative of natures physical
behavior for contacting surfaces of some classes
of materials with certain stiffness, geometry and
frictional characteristics.

OVERCOMING CONVERGENCE &

VALIDATION OBSTACLES
The 7-Step FEA System
7-Step FEA is a structured system for
performing and validating finite element analysis
in engineering design developed and presented
by the author for customized training. A
flowchart of the upper level categories in the
decision tree is shown in the Appendix.
Two, nested iteration loops within the flowchart
focus on 1.) Computational models and 2.)
Project objectives.
1.) Model and solution parameters are
evaluated against the idealization to

Figure 9 2D coupled thermo-mechanical study

of molten weld deposition and freezing., 300x
7-Step FEA was conceived as an educational
vehicle to nurture rapid, competent growth of a
companys core resource, the engineer/analysts
performing advanced FEA. Hinton [2] described
a basis for the four computational model
components, steps 2 5, in 1992.
Just as in engineering design, the optimization of
the process of FEA cannot be effectively
undertaken until a complete, technical
description or model is generated which
clearly defines how such analysis work is or
should be organized, performed and delivered.
This model should consider how it affects and is
affected by its component parts the analyst,
client (or management), data, software,
hardware, physical product or process and the
project environment.
Thus, the 7-Step FEA system was developed
as to model the FEA project process, a structured
and professional set of decision sets and
guidelines for performing and validating all
linear, nonlinear, static and dynamic, steady state
and transient physics-based events comprising
most main stream types of FEA.

This system provides an organized framework

for the many decisions required of the analyst.
Initially, it may be used as a checklist to help
plan the primary tasks for an FEA project. This
framework could then be custom-fitted with
guidelines and techniques specific to the
industry, product and analysis types required to
achieve convergence and facilitate validation.
The structure inherent in 7-Step FEA is rarely
found in such specialized fields as engineering
analysis, and the author hopes that NAFEMS and
other relevant institutions will consider
incorporation of this system model into their
global training programs.

Melting Support Techniques

Variable stiffness components can be added to
the assembly simulation model in order to
provide a temporary level of support and stability
control for component contact interactions.
Figure 10 below presents an assembly model that
utilized melting supports or tabs between the
cover and housing, as well as a common one
joining all bolt heads together.

be taken to consider the rather large strains

experienced as energy is absorbed, so not to
develop invalid elements.
They may be added between assembly model
parts experiencing initial surface contact for
instance. Or, they may be added for initially
fixing a component in space. Their stiffness will
ultimately be removed, gradually after the
temporary stability objectives are achieved.
Such stiffness control can be simply afforded by
use of thermo-elastic material property
definitions. This inspired the label melting
support as an artificial temperature rise is used
to reduce and remove their stiffness.
The melting supports themselves do not require
birth and death states a capability available
in some FEA codes. Instead, they simply remain
connected to one of the parts being stabilized.
Relative to stiffness matrix compilation, the
model components connected and stabilized by
these melting supports actually become part of
the same stiffness system and behave as one part,
temporarily.
The above technique can provide support and
stability for any duration. These supports are
controllably melted as slowly as needed.
Thus, the convergence problems associated with
initial surface contact requiring solution
convergence on the very first time step are
overcome.
Material damping features may also be added to
these supports if energy dissipation is sought.
One version of the featured ball valve assembly
model also included a melting support fashioned
as an internal annulus portion of the blind hole.
Preloading of thread axial and radial strains
normally absorbed on the very first time step
were thus relaxed to facilitate convergence.

Super Modeling Technique

Figure 10 Drive assembly model with

preloaded bolts stabilized by melting supports
at the bolt head and between cover and housing.
These support components are comprised of new
continuum elements attached to the target parts
and sharing common nodes. Their shape may
take any convenient form. However, care must

Super models were first introduced by the

author at a NAFEMS seminar presentation [3],
and represent an extremely efficient method of
both achieving convergence while addressing
validation.
In general, a super model as opposed to a super
element is a collection of various types of
elements into a functional model component,
such as a bolt connected to an annular portion of
its blind hole.

The super model of a bolt and hole was initially

tested, refined, duplicated and finally inserted
into the featured ball valve model. However, its
status as a complete model allowed independent
processing of the preload condition solution,
including surface contact loading and melting
supports.
A super models size and processing time are
always fractional relative to the assembly model.
Thus, experimentation on super model behavior
is possible. This fact is critical in light of relying
upon its proper function, such as preloading, in
the assembly.
Experimentation of supermodel behavior
includes all solution parameters that prepare the
analyst to predict numerical convergence
behavior and use appropriate settings for the
assembly.
Welding is another formidable use of super
models, as the modeling of welding processes is
extremely challenging without considering the
huge geometric burden of the joined materials.

participating nodes.
However, unstructured
meshes may leave this possibility somewhat to
chance.
The most common error is simply the
specification of nodal contact stiffness that is too
soft or too hard. The stiffer material should
generally be used to calculate the stiffness.
Contact stiffness is also proportional to the
contact distance modeled between the
participating parts. Thus, this parameter is used
in calculating the nodal contact stiffness
mentioned above.
Bilinear contact stiffness is also available and is
most valuable when gapping is not required. An
initial soft stiffness is used to establish contact
on the first time step, actually representing a
gapped condition. Then, as the parts attempt
penetration, stiffness is increased, representing
the realistic material contact stiffness.

Surface Contact Studies

Individual test model studies of less than five or
ten elements should be conducted as a matter of
course for all analysts aspiring to assembly
simulations involving surface contact.
Such studies should address the intended contact
of a single node upon another elements edge
(2D) or surface (3D). It should be possible to
predict the exact magnitude and reaction force
vector experienced at any point in time.
NAFEMS benchmarks by Feng and Prinja [4]
should be studied, along with Hertz contact
behavior which affords an exact analytic solution
(Figure 11).
The author has performed several in depth
studies of Algors surface contact capability,
implemented using a penalty stiffness approach.
Results indicated the expected outcome of
convergence difficulties at nodes aligned with
contacting element boundaries.
Recent testing [5] indicated vast improvement in
surface contact convergence and predictable
behavior resulting from implementation of the
authors suggestion to allow marginal extension
of surface contact boundaries.
Key to surface contact success for assembly
simulations is the nonalignment of all

Figure 11 Hertz contact studies for bearings

and sliding over retainer features. Subsurface
peak stresses govern case hardening decisions.
Mesh density should be quite uniform for each
surface so defined, as each participating node in
contact will contribute its normal reaction load.
Thus, for surface-to-surface contact, as opposed
to node-to-surface contact, nodes on both
surfaces are contributing stiffness.
Definitely, initial contact surface interference is
the most significant technique to facilitate
convergence for surface contact solutions. There
is absolutely no reason to process an impact
event when such dynamics generate no valuable
data.

Precision Surface Contact

Precision surface contact goes beyond the
normal expectations of simply transferring gross
body loads from one surface to another, it can
reflect realistic sliding, load redistribution,
damping and frictional resistance.
Gearing is a perfect example of the need for
precise surface contact, especially for analysis of
dynamic gearing events such that depicted in the
extreme gearing example shown in Figures 12
& 13. This is a hypothetical gear set of five and
ten teeth having approximately a 1.0 contact
ratio, a parameter known to initiate and support
damaging gear dynamics.
Figure 13 Dynamics of an accelerating
extreme gear set (Figure 12), evidencing nonconstant velocity due to stiffness variations,
profile error and surface roughness, sensitized by
its design with a contact ratio of 1.0.

Contact Interference Techniques

If possible, all surface contact should be
established as an initial interference. The key
issue here is that the value of this interference
can be on the order of 1.0E-6 inch or so!

Figure 12 Nonstandard 5-tooth pinion and 10tooth gear having 43 degree pressure angle.
The surface contact contours of the mating parts
inherently contain a piece-wise linear surface
roughness feature. This parameter can be
manipulated primarily via mesh adjustments to
investigate surface finish variation effects.
Important to recall and to have studied, as in the
surface contact testing discussed earlier, that
surface contact elements may have discontinuous
stiffness (=0) at convex surface element
boundaries.
At concave surface element boundaries, stiffness
may be duplicated such that stiffness of 200% 800% or more may theoretically exist, given that
all surface elements are triangular and never
quadrilateral.

Essentially, since convergence is facilitated by

the smaller values of nodal displacement or
rotation, it is wise to plan interference parameters
with this in mind. The author typically places
contacting parts 0.001 or less apart in model
space, and provides for interference of only 0.1%
of this value.
Validation can be facilitated by such tight
clearance of a models contacting surfaces, as it
allows for larger magnification during postprocessing without undue distortion (due to
motion across the gap). This supports visual
verification of otherwise undetectable part
contact behaviors.
Figures 14 & 15 present nonlinear dynamic
results of a gas compressor assembly model
utilizing such tight clearances (except in one
region.)
Interference schemes are routinely used in
conjunction with melting support techniques.
For assembly models, control over the relative
motion of each part after initial contact, can also
facilitate convergence. Fewer nodes moving
shorter distances are always better.

The most detailed approach for the sliding would

be to model the shank surface contact with the
hole ID, but this is not usually necessary.
An efficient technique to overcome both
problems is to use one or more embedded beam
elements of low stiffness connected to the
bottom of the blind hole. All but the axial
rotation and transverse translation degrees of
freedom can be maintained to restrain spinning
and sliding.
Ultimately, the elastic deflections of both mating
parts will develop a pair of annular dish-shaped
surfaces that fit one into the other, thus,
nominally restraining the sliding tendency.
Orthotropic melting supports located to fit as a
wrench could also be used to limit rotation while
a much softer directional stiffness allows for the
bolts axial stretch.

Figure 14 Gas piston assembly simulation

reciprocating at 600 rpm with bolt rod tension,
pressure and thermal loading, and no friction at
contacting surfaces, Mises stresses, 25x

Frictionless Stability Techniques

Due to the problematic nature of frictional
surface resistance, it is highly recommended to
omit this simulation feature from the initial
assembly model solutions.
Instead, super models or other test models are a
better venue for investigating this capability,
until ultimately, the project objectives may
require that it be included.
The ball valve model featured in this paper was
processed with no surface friction. However,
incremental bolt head shifting was noted to
hamper convergence once or twice during small
periods of the events duration. Increased mesh
refinement under the bolt head could have
improved this phenomenon.
Frictionless bolted connections can be quite
unstable during preloading without friction,
attempting to spin on their axes or slide off of the
clamped part. Melting supports, detailed earlier,
can resolve this issue, but only until their
stiffness are melted away.

Figure 15 Cyclic loading develops these

stresses in the concave fillets of the piston
halves.

Over-damping Techniques
For nonlinear dynamic simulations, such as the
gas piston model discussed earlier, a temporary
over-damping of the event can be imparted to

help stabilize the convergence of a difficult

solution. Then, a restart analysis would be
processed with either no damping or an
acceptable level to meet the FEA project
objectives.

time due to less nodes and lower convergence

error values. Higher order elements can then be
used in the final solutions if needed.

Figure 15 below depicts results from model

testing to validate the implementation of Raleigh
damping within the Algor software for a simply
supported reinforced concrete beam.

Transition elements include those having less

than a full compliment of mid-side nodes at each
edge of the element. This allows for nodal
compatibility with higher order elements on one
side and lower order elements on the other side
of the transition element set.

Transition Element Meshing

Judicious use of higher order element by

combining them with lower order and transition
elements will permit the processing of the most
accurate model for a given degree of freedom
budget.
However, such transition elements are not
universally supported for this use. The author
appeals to the FEA vendors for such tools.

Relative or Hierarchal Model Feature

Validation
Relative or hierarchal forms of validation are
possible for each model feature.
This is
especially valuable when having access to a
companys FEA model database.

Figure 16

Load Curve Refinement Techniques

The relative approach is one of relating project

model features to validated, established model
features.
Basically, geometric, elemental,
constraint set and solution parameter model
features are individually validated using small
models. These validation models are then
modified by one feature at a time as a control
parameter.
This captures quantitative
relationships for the common feature variations.

It is worth noting that even the use of 1000 point

sinusoidal load curves will not guarantee smooth
harmonic motion if the numerical values are
truncated to just a few significant digits.

Such feature-based validation may be similar to

the lab testing of nonlinear plastics or rubber,
whereby characteristic modes of loading for
specific constraint sets will yield predictable
behavior.

Focused testing on this topic confirmed that upon

automatic time step reduction during the
processors attempt to achieve convergence the
incremental changes in curve slope excited
model natural frequencies!
Such spurious
harmonics are to be avoided at all cost.

Hierarchal variations of the relative model

feature validation approach involve parent-child
features and are not as broadly applicable as the
relative approach.

Incompatible Mode Element Use

Use of incompatible mode or nonconforming
elements can often generate results as accurate as
higher order elements, with much less processing

A set of model geometry, mesh density, element

features and other parameters will generate
similar results for similar loading and constraint
sets.

CONCLUSIONS
Advanced, nonlinear FEA has been costeffectively used to model assembly simulations,
some of the most demanding but valuable
analysis opportunities facing engineers today.
Some of the most common and offensive
spurious behaviors found to defeat nonlinear
FEA solution convergence have been identified.
General guidance and novel techniques have
been described for overcoming such obstacles.
As well, FEA model and results validation has
been labeled as the process to confirm, establish
and help quantify FEA value. Such efforts are
notably difficult.
Several approaches for
accomplishing such validation have been
described.

REFERENCES
1.

Marchand, L., Laviolette, D., Derenne,

M., NPS 4 Class 150 Bolted Flanged
Joints Subjected to Pressure and
External Bending Loads, WRC Bulletin
450, April 2000, pg 18.

2.

Hinton, E., NAFEMS Introduction to

Nonlinear Finite Element Analysis,
1992, Bell and Bain Ltd, Glasgow, pp
9-14.

3.

Orgeron, K J Validation of FEA

Optimization Projects: Cost Effective
Procedures, presented at NAFEMS
Seminar: Advances in Optimization
Technologies for Product Design,
October, 2001, Chicago, pg 9.

4.

Prinja, N K and Feng, Q Benchmark

Tests for Finite Element Modeling of
Contact, Gapping and Sliding, Proc.
NAFEMS 2001 World Congress, April,
Lake Como, Italy, pp 1005-1016.

5.

Orgeron, K J Cover letter of Draft

Report:
Contact Surface Extension
Testing, Integra Engineering, Inc.,
December 5th, 2001, pp 1-30.

APPENDIX